Research ArticleMA Cation-Induced Diffusional Growth of Low-Bandgap FA-CsPerovskites Driven by Natural Gradient Annealing

Taiyang Zhang,1 Yuetian Chen ,1 Miao Kan,1 Shumao Xu ,1 Yanfeng Miao,1

Xingtao Wang,1 Meng Ren,1 Haoran Chen,1 Xiaomin Liu,1 and Yixin Zhao 1,2

1School of Environmental Science and Engineering, Shanghai Jiao Tong University, 800 Dongchuan Rd., Shanghai 200240, China2Shanghai Institute of Pollution Control and Ecological Security, Shanghai 200240, China

Correspondence should be addressed to Yixin Zhao; yixin.zhao@sjtu.edu.cn

Received 21 April 2021; Accepted 4 July 2021; Published 18 August 2021

Copyright © 2021 Taiyang Zhang et al. Exclusive Licensee Science and Technology Review Publishing House. Distributed under aCreative Commons Attribution License (CC BY 4.0).

Low-bandgap formamidinium-cesium (FA-Cs) perovskites of FA1-xCsxPbI3 (x < 0:1) are promising candidates for efficient and robustperovskite solar cells, but their black-phase crystallization is very sensitive to annealing temperature. Unfortunately, the low heatconductivity of the glass substrate builds up a temperature gradient within from bottom to top and makes the initial annealingtemperature of the perovskite film lower than the black-phase crystallization point (~150°C). Herein, we take advantage of suchtemperature gradient for the diffusional growth of high-quality FA-Cs perovskites by introducing a thermally unstable MA+ cation,which would firstly form α-phase FA-MA-Cs mixed perovskites with low formation energy at the hot bottom of the perovskitefilms in the early annealing stage. The natural gradient annealing temperature and the thermally unstable MA+ cation then lead tothe bottom-to-top diffusional growth of highly orientated α-phase FA-Cs perovskite, which exhibits 10-fold of enhancedcrystallinity and reduced trap density ( ~ 3:85 × 1015 cm−3). Eventually, such FA-Cs perovskite films were fabricated into stablesolar cell devices with champion efficiency up to 23.11%, among the highest efficiency of MA-free perovskite solar cells.

1. Introduction

Organic-inorganic hybrid metal halide perovskites withsuperior photovoltaic (PV) performances have drawn enor-mous research interest within the last few years [1–3]. Thepower conversion efficiencies (PCEs) of the fabricated perov-skite solar cells have dramatically progressed from the unsta-ble 3.8% in 2009 to a certified record of 25.5% in 2020 [4–6].Beyond the high efficiency, more and more concerns on thelong-term stability have risen. Therefore, FA-based perov-skite with enhanced thermal stability and a narrower band-gap is a promising candidate to pursue robust and efficientsolar cells [7–11]. However, the photoactive black phase (αphase) usually requires higher annealing (~170°C) and tendsto transform into the inactive nonperovskite yellow phase (δphase). To lower the α-phase crystallization temperature andstabilize the black phase, the introduction of smaller cationssuch as MA+ and Cs+ or anions like Br- to form mixed catio-n/anion alloy perovskites has been one of the most popularapproaches [8, 12–14]. However, such alloying procedure

needs to be carefully tuned as the incorporation of Br- andsmaller cations leads to broadening of the perovskites’ band-gap and raises concerns on phase segregation [6, 15–20].

The annealing route is another important but often lessconcerning factor for the fabrication of high-quality FA-based perovskites [21–24]. The regular hotplate thermalannealing process is the most adopted approach for large-scale perovskite deposition. The naturally formed bottom-to-top annealing temperature gradient exists when the glasssubstrate is placed on the hotplate for thermal annealingdue to the low heat conductivity of glass, as shown inFigure S1 [25]. According to previous studies [24], itusually takes several seconds for the entire perovskiteprecursor film to reach the set temperature. Suchunexpected phenomenon may be crucial for a temperature-sensitive process. As the crystallization of FAPbI3 issensitive to the annealing temperature, such temperaturegradient might manipulate the FA perovskite ofcrystallizing into the δ phase with low formation energy inthe initial annealing stage, leading to an inevitable δ-α-

AAASResearchVolume 2021, Article ID 9765106, 11 pageshttps://doi.org/10.34133/2021/9765106

phase transformation and deterioration of the quality of thefinal perovskite films [12, 26]. Meanwhile, Pool et al. foundthat radiative thermal annealing could produce high-qualityFAPbI3 films because of the faster and more uniformheating flux direction [23]. However, to the best of ourknowledge, few research efforts have been invested inutilizing this gradient phenomenon.

In this work, we report the diffusional growth of high-quality low-bandgap FA-Cs perovskites via an intermediateengineering strategy utilizing the annealing temperature gra-dient. We found that the extra MA+ cation in the FA-Csperovskite precursor could significantly lower the formationenergy of α-phase perovskite by forming FA-MA-Cs perov-skite seeds firstly at the bottom of the perovskite films. Drivenby the thermal gradient, such α-phase perovskite crystalseeds could diffuse efficiently, thus promoting the bottom-to-top gradient growth of perovskite and suppressing thecrystallization of the unwanted δ phase. Finally, a high-performance MA-free FA0.95Cs0.05PbI3-based solar celldevice with a PCE of over 23% can be obtained. Such valueis among the highest efficiency of the MA-free perovskitesolar cells. Besides the high efficiency, the stability of thedevice is also greatly enhanced, which could retain 90% ofits initial PCE after 1000 h white LED light soaking.

2. Results

We chose the FA0.95Cs0.05PbI3 perovskite as the lightabsorber layer because of its excellent thermal stability and nar-row bandgap (~1.53 eV) [27]. In our intermediate engineeringprocess (denoted as IE), different amounts of MAI were addedinto the FA0.95Cs0.05PbI3 precursors to form a new IE solution(see methods in Supplementary Materials for experimentaldetails). We firstly compared the effect of the new precursoron the final film quality. As shown in Figure 1(a), both theIE sample and the control sample show the same absorptiononset at ~813nm, indicating the successful formation of~1.53eV bandgap perovskites. The baseline of the controlsample in the UV-vis spectrum is much higher than thatof IE, indicating a rougher surface of the control sample. Itis worth noting that the two samples had exactly the sameabsorption onset as the previously reported FA0.95Cs0.05PbI3perovskite [28]. The photoluminescence (PL) spectra of theIE sample and the control sample (Figure S2) show a peakat the same position (~810nm) while the PL intensity ofthe IE sample is much stronger than that of the controlsample. Furthermore, NMR and TGA shown in Figure S3and S4 excluded the existence of MA+ residues in the finalfilm, confirming that the IE sample is a pureFA0.95Cs0.05PbI3 perovskite.

Figure 1(b) shows the XRD patterns of the IE and controlsamples. The IE sample exhibits two strong and sharp peaksat ~13.9 degrees and 28.1 degrees, which could be ascribed tothe (100) and (200) plane signals of α-phase FA0.95Cs0.05PbI3perovskite. There are no peaks belonging to the δ phase orMAPbI3 or impurities in this XRD pattern (Figure S5),indicating a highly crystalline and phase-pureFA0.95Cs0.05PbI3 perovskite film. The peak intensity of theIE perovskite films is almost 10 times stronger than that of

the control sample. Synchrotron radiation grazing-incidencewide-angle X-ray scattering (GIWAXS) analysis is thenperformed to probe the crystal orientation in the perovskitefilms. As revealed in Figure S6, the control sample shows aweak diffraction ring, indicating the random crystal planestacking. As for the IE sample (Figure 1(c)), a dark reddiffraction mottling could be found at qz = 10 nm−1, clearlyrevealing the strong (100)-oriented growth of theFA0.95Cs0.05PbI3 films. The azimuth angle plots inFigure 1(d) show the integrated intensity plots of the q = ~10 nm–1 peak. Compared to the overall low intensity of thecontrol sample, the IE sample demonstrated a sharp peak at~90 degrees, which further strengthened the preferred (100)orientation of the IE sample [29–31].

SEM images of the IE and control samples are shown inFigures 1(e) and 1(f) and Figure S7. The IE sample exhibitsa compact surface and clear grain boundaries with >500nmgrain size, while the control sample shows pinholes with~200nm grain size. Moreover, the cross-sectional SEMimage of the IE sample (Figure 1(f) inset) shows a quasi-single crystalline feature in the vertical cross-section whilethe control sample shows polycrystalline grains with lots ofvoids between the film and the substrate in the verticaldirection. The IE films with larger grain size, bettersubstrate contact, and out-of-plane orientation are morefavorable for the fabrication of high-efficiency perovskitesolar cells, which will be discussed in detail later. The IEsample also demonstrates excellent thermal stability, asshown in Figure S8; the sample shows no visibledegradation after aged at a ~100°C hotplate for 10 days.Here, we should note that the amount of excess MA+

cations needs to be carefully tuned for a nice film. Wefound the 0.2MAI sample as the optimal addition amountwhile too much or less MA+ cations would deteriorate thefilm quality, and the 0.2MAI sample demonstrates the bestperformance (Figure S9–S11). Therefore, we chose the0.2MAI recipe for further discussion and optimization.

The above results confirmed that intermediate engineer-ing could greatly change the film quality; we then exploredthe role of the extra MA+ cation in the crystallization process.Firstly, the as-prepared precursor films, which were annealedat 60°C for 30 s to repel the solvent, were examined. As shownin Figure S12, the IE and control precursor films show asignificant difference in color at the initial stage; the IEsample is brown while the control sample is yellowish. UV-vis spectra in Figure 2(a) reveal that only the δ-phaseabsorbance feature (~400nm) is observed in the controlsample, while the IE film has a distinct shoulder peak at~763nm and a δ-phase-related absorbance peak at~400nm. It is highly likely that the phase segregation tookplace immediately in the IE precursor film with theformation of both α-phase perovskite seeds and δ-phaseperovskites. It is also worth noting that such α-phaseperovskite seeds have a wider bandgap than the targetperovskite, indicating that the precrystallized perovskiteseeds could contain some MA+ cations. The XRD patternof the IE precursor film in Figure 2(b) exhibits one strongpeak at ~11.7 degrees and another three peaks at ~13.9,28.1, and 42.2 degrees, which should be assigned to the

2 Research

δ-phase and α-phase perovskites; no MAPbI3 peaks areobserved (Figure S13), respectively [32–35]. In contrast,only a very sharp δ-phase peak could be observed in thecontrol precursor film. The AFM images in Figure S14show that the control precursor film has a smaller grainsize of 200~500nm with clear grain boundaries while thegrain boundaries of the IE precursor sample are lessdistinct, confirming that the crystallinity of the IE sample

is weaker than that of the control sample. Both the UV-vis absorption and the XRD results suggest that the extraMA+ cation in the IE sample may induce theprecrystallization of possible α-phase FA-Ma-Cs mixedcation perovskite seeds and cause the phase segregationbetween the δ phase and the α phase. Additives of MAIand MACl show a similar effect in tuning crystallizationdynamics.

600 700 800 900

Abs

Wavelength (nm)

ControlIE

Inte

nsity

(a.u

.)

10 20 30 40 502theta (degree)

(100)(200)

ControlIE

IE

0105–5

qxy

(nm–1)

qz (n

m–1

)

–10 0

100

200

300

400

500

0

5

10

15

20

25

0 30 60 90 120 150 180

Azimuth (degree)

(f)(e)

(d)(c)

(b)(a)

Inte

nsity

(a.u

.)

ControlIE

1 𝜇m

Control

500 nm

1 𝜇m

IE

500 nm

Figure 1: Effect of the intermediate engineering process on the quality of perovskite films. UV-vis spectra (a) and XRD patterns (b) ofannealed films with and without IE. GIWAXS pattern of the IE sample (c). Radially integrated intensity plots of the q = ~ 10 nm–1 peakfrom the 2D GIWAXS patterns (d). SEM images of annealed films of the control (e) and IE (f). Corresponding cross-sectional SEMimages are shown as inset; the voids between the film and the substrate are marked by red circles.

3Research

Since, in principle, regular UV-vis and XRD analysesyield statistical results, we could not get the spatial distribu-tion of the α phase in the films from these characterization

techniques alone. We then performed a double-side PL test,which allows us to determine the existence of inhomoge-neous growth of perovskite in the films. As illustrated in

𝛿 phase

400 500 600 700 800

Abs

Wavelength (nm)

ControlIE

𝛼 phase

10 20 30 40 502theta (degree)

: 𝛿 phase: 𝛼 phase: Substrate

IE

Control

Inte

nsity

(a.u

.)

Em. Em.

Perovskiteside

Glassside

Ex. Ex.

650 700 750 800 850

0.0

0.5

1.0

Wavelength (nm)

Nor

mal

ized

PL

Perovskite sideGlass side

IE sample

Sn2+

Pb2+

I–

Cs+

MA+

FA+

Sputter time (s)

(e)

(c) (d)

(a) (b)

Inte

nsity

(cou

nts)

Sn2+

Pb2+

I–

Cs+

MA+

FA+

Perovskite layer Substrate

1000 200 300 400 500

Figure 2: Characterization of the IE and control perovskite precursor films. UV-vis spectra (a) and XRD patterns (b) of the control and IEprecursor films. Schematic diagram of the double-side PL test (c). PL spectra (d) measured from both sides of the IE precursor film. ToF-SIMSdepth profile of the IE precursor film (e).

4 Research

Figure 2(c), in this test, the PL spectrum collected from theperovskite side is dominated by signals from the top surfaceof the perovskite film, while the PL spectrum measured fromthe glass side gives more information on the perovskite adja-cent to the glass substrate [36]. As shown in Figure 2(d), thePL spectra collected from two sides of the IE sample show asignificant difference: the PL peak collected from the glassside has blue-shifted from the perovskite surface side, indi-cating that the perovskite seeds in the bottom part have awider bandgap and could be more MA-rich than the topones. As for the control sample (Figure S15), the weak PLsignals exhibit no variation in peak positions between thetwo sides. Time-of-flight secondary ion mass spectrometry(ToF-SIMS) depth analysis is performed to further explorethe existence of a possible compositional gradient. Asshown in Figure 2(e), the signals of Cs+, FA+, and Pb2+ areuniformly distributed across the entire depth of theprecursor film, while the signal of MA+ is slightly strongernear the bottom than the surface. This disparity indistribution confirms that MA+ is richer near the bottom,which is well consistent with the PL results. Here, we couldpropose a hypothesis that the diffusional growth of α-phaseperovskite happens across the vertical direction in the IEsample. The precrystallized α phase near the bottom ismore MA-rich than that near the surface because a higheramount of smaller MA+ cations can help to lower theformation energy barrier for the α phase [37, 38]. Suchdiffusional growth then induces the oriented crystallizationof high-quality perovskite films, which will be furtheranalyzed later.

These material characterization results have proven thecoexistence of the α phase and the δ phase in the IE precursorfilm, and the α phase was gradient-distributed along the ver-tical direction. As the annealed final film shown inFigure 1(b) is a pure α phase, we then investigated thedetailed crystallization process of the IE samples to under-stand the δ-α-phase transformation. As shown inFigure 3(a), the IE perovskite film’s absorbance onset red-shifted from ~801 nm to 813 nm after annealing at 150°Cfor ~30 s, which is very close to the feature of the final film.Further annealing duration led to the enhancement of absor-bance intensity instead of onset shift, indicating a fast-diffusional perovskite phase crystallization through thewhole film.

The PL analysis in Figure 3(c) and Figure S16 also showsthat after the initial high-temperature annealing, thedifference of PL peak positions between the two sidesvanished. This observation further confirmed thehypothesis that the MA+ cation-contained perovskite seedswould diffuse quickly from bottom to top. The XRDpatterns shown in Figure 3(b) also indicate that after 30 s ofannealing, the α-phase perovskite peaks still existed whilethe δ-phase peak at 11.6 degrees completely disappeared.Moreover, the intensity of perovskite peaks has greatlyenhanced after 35min of annealing. The morphologyevolutions characterized by AFM imaging (Figures 3(d)–3(f))demonstrate similar results. The grain size started todecrease while the grain boundaries became much clearerwith annealing, suggesting that the phase and component

transformation occurred during the annealing process. Thecontrol samples show a quite different evolution process asthe UV-vis absorption spectra and XRD patterns revealedin Figure S17 and S18. The typical α-phase absorbancestarts to rise after 5min of annealing, and then theabsorbance intensity slowly reaches the maximum valuesafter 10min of annealing. The XRD patterns only showstrong δ-phase peaks in the initial annealing, and then theα-phase peaks start to emerge after 5min of annealing, butthe δ-phase peaks still coexist. The δ-phase peaks vanishedafter 10min of annealing. Such results are well consistentwith the UV-vis spectral profiles, further confirming themuch slower phase transformation in the control samples.

Based on all the above experimental results, we finallypropose a growth mechanism of our IE method. As illus-trated in Figure 3(g), the diffusional growth process couldbe divided into two main steps: the short time step I (~30 s)and the much longer step II (~35min). In step I, DMSO-containing PbI2-DMSO adducts form after the solvent engi-neering process, which would retard the further crystalliza-tion process. As we know, DMSO has a strong capability ofcoordinating with PbI2 and MAI but weakly interacting withFAI, as evidenced by Figure S19 [19, 20, 39]. As thedissociation rate of DMSO is faster via annealing, DMSOmolecules close to the hotter bottom are firstly dissociatedupon annealing, triggering the crystallization process. Theformation of the temperature gradient mainly impacts thisstep. In the actual experimental settings, the initial filmtemperature is lower than the value set on the hotplate.This temperature gradient built up inside the perovskitefilm during thermal annealing leads to the formation ofyellow-phase perovskite. Previous works have confirmedthat the formation energy plays an important role in thecrystallization process of perovskites [38, 40]. The yellow δphase is more favorable for FA1-xCsxPbI3 (x < 0:1) whenthe annealing temperature is below the phase transitionpoint (~150°C) [12]. In our intermediate engineeringmethod, the formation energy of black-phase perovskitewas greatly reduced by first the formation of FA-MA-Csperovskite [37, 38]. Upon annealing, the hot bottom helpedto dissociate the DMSO molecule and precrystallize theblack-phase perovskite seeds in the IE sample. Then, as theannealing proceeds, the crystallization of black-phaseperovskite diffused vertically to the top of the film. In suchbottom-to-top diffusional growth of FA-Cs perovskite, wecould successfully utilize the annealing temperaturegradient and the thermally unstable MA-based perovskiteseeds. In contrast, if the precursor film is annealed from thetop side via hot airflow, the perovskite film shows much-reduced crystallinity than that of the bottom-annealedsample (Figure S20). What is more, the intensity ratio ofthe peak located at ~14 degrees ((100) plane) and ~20degrees ((110) plane) was also much smaller than that ofthe bottom-annealed samples, indicating the less orientedgrowth of the perovskite film. Such results furtherconfirmed that such temperature gradient is an importantfactor for the preparation of highly oriented perovskitefilms. In step II, as the interaction of PbI2 and FA+ is muchstronger than that of MA+, the MA+ cation was replaced by

5Research

the FA+ and then repelled from the film to further improvethe film quality (Figure S21). As evidenced by previousreports [41, 42], such process may last for tens of minutes.Although the duration of step I is relatively short, itsimportance cannot be ignored. The MA+ cation-inducedprecrystallized perovskite seeds at the bottom in step I

could also act as templates for the further growth ofperovskite films as they could relieve the strain built duringthe δ-α-phase transformation. As we know, the δ phaseconsists of face-sharing PbI6 octahedra while the PbI6octahedron in α-phase perovskite is corner-sharing. Theface-sharing mode is more compact than the corner-

600 700 800

AnnealedAb

s

Wavelength (nm)5 s30 s35 min

Inte

nsity

(a.u

.)

10 20 30 40 50

35 min

30 s

5 s

2theta (degree)

(g)

(f)(d)

(b) (c)(a)

5 s 30 s 35 min760

770

780

790

800

810

Annealing stages

PL p

eak

posit

ion

Glass sidePerovskite side

1 𝜇m

200 nm

0 nm

5 s

(e)

1 𝜇m

200 nm

0 nm

30 s

1 𝜇m

200 nm

0 nm

35 min

Perovskiteseed diffusion

𝛼-phase perovskite

𝛿-phase perovskite

(I) (II)

MA+ repel

Figure 3: The evolution process of the IE perovskite films within annealing. UV-vis spectra (a), XRD patterns (b), PL peak positions of thisdouble-side test (c), and AFM images (d–f) demonstrating the evolution of the precursor films with different annealing durations. Schematicillustration of the proposed seed diffusion-assisted crystallization of the perovskite films (g).

6 Research

sharing mode; hence, an expansion of the PbI6 octahedronwould take place in this phase transformation process [26].Such feature is somehow similar to the often-considereddifficult “two-step” method, which involves thetransformation from highly crystalline PbI2 to MAPbI3[43–47]. For the control sample, the highly crystalline δphase is so hard to corrode, leading to the smaller crystalsize and formation of pinholes in the films [26]. Thisdifficult δ-α-phase transformation also caused more criticallattice tension/compression within the perovskite film,which could lead to more serious strain [48]. Such strainscould induce more ion migration and generate more defectsites, which not only deteriorated the device performancebut also reduced the stability of perovskite films [49, 50].On the contrary, the crystallinity of the δ phase in the IEsample was greatly reduced by the perovskite seeds. Thus,the expansion of the PbI6 octahedron was promoted andthe lattice stains were relieved, which finally leads to thegrowth of highly crystalline, well-oriented, pinhole-free,and robust perovskite films.

The impact of our IE method on the charge transferdynamics was further investigated. Time-resolved photolu-

minescence (TRPL) results are shown in Figure 4(a), andthe PL lifetime could be obtained by biexponential fittingusing the following equation:

Y = A1 exp −t/τ1ð Þ + A2 exp −t/τ2ð Þ: ð1Þ

The IE sample shows a much longer average PL lifetime(32.81 ns) than the control sample (5.15 ns), indicating thatthe trap state-assisted nonradiative recombination is greatlysuppressed in the IE sample [51]. The electron-only devicesusing the FTO/TiO2-SnO2/perovskite/PCBM/Ag structurewere also fabricated to evaluate the trap density of the IEand control samples. As shown in Figure 4(b), the linear sec-tion (blue line) at low bias voltage indicates the ohmic-typeresponse, the trap filling region is marked by the light cyanline, and the trap filled limit voltage (VTFL) lies in the kinkpoint of the two regions. The trap density (N t) of perovskitescan be calculated by the following equation:

N t =2ε0εrVTFL

eL2, ð2Þ

0 100 200 300 400

10–3

10–2

10–1

100

Decay time (ns)

ControlIE

Nor

mal

ized

PL

inte

nsity

10–2

10–1

100

101

102

0.01 0.1 1

ControlIE

Curr

ent (

mA

)

Voltage (V)

VTFL: 0.14 V

VTFL: 0.25 V

0 200 400 600 800 10000.0

0.2

0.4

0.6

0.8

1.0

Time (𝜇s)

ControlIE

Nor

mal

ized

pho

tovo

ltage

ControlIE

0 200

Nor

mal

ized

pho

tocu

rren

t

400 600 800 1000 1200

0.0

0.2

0.4

0.6

0.8

1.0

Time (𝜇s)

(d)(c)

(b)(a)

Figure 4: Charge transfer properties of the IE and control samples. TRPL curves (a) of the IE films and the control films. Dark I-V curves (b)of the IE- and control-based electron-only devices. TPV (c) and TPC (d) curves of the IE and control devices.

7Research

where the ε0 and εr represent the vacuum permittivity andthe relative dielectric constant of perovskites (ε = 62:23 forFA0.95Cs0.05PbI3) [28], e is the electron charge, and Ldenotes the thickness of the films. The trap density forthe control sample is ~ 6:88 × 1015 cm−3 while the IE sam-ple shows a much-reduced value of ~ 3:85 × 1015 cm−3.Such remarkable lower trap density should be benefitedfrom the larger grain size and the well-oriented growthof IE samples. The transient photovoltage (TPV) decayresult in Figure 4(c) shows that the IE sample has a longer

charge-carrier lifetime than the control sample, indicatingthe lower undesired carrier’s recombination rate of theIE sample, which is consistent with the TRPL result.Moreover, as shown in Figure 4(d), the shorter lifetimeof the IE sample in the transient photocurrent (TPC)decay curves also confirmed that the IE sample had amuch faster photocarrier transit rate than the control sam-ple. All the results further strengthen the fact that thehigh-quality IE film can reduce the defect density andfacilitate the carrier transfer and extraction [49, 52].

0 150 300 450 600 750 900 10500

0.2

0.4

0.6

0.8

1

ControlIE

Nor

mal

ized

PCE

Time (h)

(e)

(c)

(a) (b)

(d)

200 400 600 800 10000

5

10

15

20

25

Time (s)

Curr

ent d

ensit

y (m

A/c

m2 )

~22.97% @0.95V

300 400 500 600 700 8000

20

40

60

80

100

EQE

(%)

Wavelength (nm)

0

5

10

15

20

25

Inte

grat

ed J

sc (m

A/c

m2 )

0.0 0.2 0.4 0.6 0.8 1.0 1.20

5

10

15

20

25

Voltage (V)Reverse scan

Voc: 1.13 V; Jsc: 25.40 mA/cm2

FF: 0.805; PCE: 23.11%

Forward scan

Voc: 1.12 V; Jsc: 25.38 mA/cm2

FF: 0.801; PCE: 22.77%

Curr

ent d

ensit

y (m

A/c

m2 )

0.0 0.2 0.4 0.6 0.8 1.00

5

10

15

20

25

Voltage (V)

Curr

ent d

ensit

y (m

A/c

m2 )

ControlIE

Figure 5: Performance of the fabricated FA0.95Cs0.05PbI3-based solar cell devices. Typical J-V curves (a) of the IE and control perovskitedevices. J-V curve (b), EQE curve (c), and stable output curve (d) of the champion IE device. Stability test (e) of the IE and controldevices. The devices were irradiated under a white LED light source (100mW/cm2) in an N2 glovebox.

8 Research

The high-quality perovskite films were then fabricatedinto perovskite solar cell devices using the FTO/TiO2-SnO2/-perovskite/Spiro-OMeTAD/Ag planar architecture. Asshown in Figure 5(a) and Table S1, the IE devices showmuch-enhanced performance with a PCE increasing from~15% to ~21% and better reproducibility. A champion IE-based device reached a PCE of 23.11% with an open-circuitvoltage (Voc) of 1.13V, a fill factor (FF) of 0.805, and ashort-circuit photocurrent density (Jsc) of 25.40mA/cm2

(Figure 5(b)), which is much better than the controlsample. The Jsc value (25.40mA/cm2) obtained from the J-V curve is well consistent with the integrated value(24.9mA/cm2) from the external quantum efficiency (EQE)curve in Figure 5(c). Negligible hysteresis was alsoobserved, and a stable output efficiency of 22.97% wasobtained (Figure 5(d)). Beyond the high efficiency, thestability of devices is also greatly enhanced. The IE devicecould retain 90% of its initial performance after 1000 hwhite LED light soaking, while the performance of thecontrol device deteriorated dramatically (Figure 5(e)). Themuch-enhanced stability of the IE device could beattributed to its better crystallinity and less trap density,relieving the strains that inhibited the defect-triggereddegradation.

3. Discussion

In summary, we developed an intermediate engineeringmethod for the growth of highly crystalline phase-pure α-FA0.95Cs0.05PbI3 perovskites. Further investigations revealedthat the precrystallization of α-phase perovskite seeds couldbe triggered by the addition of excess MA+ cations, whichlowered the formation energy of the α-phase perovskite andtook advantage of the temperature gradient effect duringthe initial annealing process. The bottom-to-top growth ofperovskite was then promoted, and the strain within thegrowth of perovskite was relieved. The obtained films havecrystallites well oriented vertically. The film quality wasgreatly improved with enhanced crystallinity and reducedtrap density, thus suppressing the nonradiative recombina-tion and promoting the carrier transfer and extraction.Finally, the high-quality α-FA0.95Cs0.05PbI3 perovskite-based solar cells exhibited much-enhanced PV performance,and the champion device reached an efficiency of over 23%with negligible J-V hysteresis. Our intermediate engineeringapproach highlights the importance of the previouslyneglected annealing temperature gradient within the anneal-ing process. Careful tuning of this gradient would be a prom-ising tactic for the deposition of low-bandgap FA-basedperovskites and other hybrid halide perovskites for higherdevice performances and scalable film deposition.

Data Availability

The data that support the plots within this paper and otherfindings of this study are available from the correspondingauthor upon reasonable request.

Conflicts of Interest

The authors declare that there is no conflict of interestregarding the publication of this article.

Authors’ Contributions

Y. Zhao designed and directed the research. T. Zhang, M.Ren, S. Xu, M. Kan, X. Wang, Y. Miao, H. Chen, and X. Liufabricated and characterized the perovskite thin films anddevices. Y. Zhao, T. Zhang, and Y. Chen analyzed the resultsand wrote the manuscript with inputs from all authors.

Acknowledgments

YZ acknowledges the support of the NSFC (Grant Nos.22025505 and 21777096), Program of Shanghai Academic/-Technology Research Leader (Grant No. 20XD1422200), andCultivating fund of Frontiers Science Center for Transforma-tive Molecules (2019PT02). TZ acknowledges the support ofthe Initiative Postdocs Supporting Program (Grant No.BX20180185) and China Postdoctoral Science Foundation(Grant No. 2018M640387). We gratefully thank Dr. TieyingYang and Dr. Wen Wen of Shanghai Institute of AppliedPhysics, CAS, for their assistance with GIWAXS analysis.

Supplementary Materials

Materials and methods. Fig. S1: (a) the temperature profileevolution measured on the top of substrates during the initialannealing process; (b) the temperature was measured using aprecise thermometer with a thermal coupler placed on thetop of substrates. Fig. S2: PL spectra of perovskite films pre-pared from the IE and control precursors. Fig. S3: 1H NMRspectra of FA0.95Cs0.05PbI3 perovskite films prepared fromour IE method. Fig. S4: TGA and the first derivative ofTGA curves for FAPbI3 at a heating rate of 10°Cmin−1 inN2. Fig. S5: enlarged XRD pattern of Figure 1(b). Fig. S6:GIWAXS pattern of the annealed control sample. Fig. S7:cross-sectional SEM images of the perovskite films preparedfrom the IE and control precursor solutions. Fig. S8: UV-visspectra of IE perovskite films before and after aged on a100°C hotplate for 10 days in an N2 glovebox. Fig. S9: com-parison of SEM images of the annealed perovskite films withdifferent MAI contents; the scale bar is 1μm. Fig. S10: com-parison of UV-vis spectra of annealed perovskite films withdifferent MAI contents. Fig. S11: J-V curves of perovskitesolar cell devices fabricated by precursors with differentMAI contents. Fig. S12: photos of perovskite films using thecontrol (a) and IE precursors (b) at the initial (top) and final(bottom) annealing stages. Fig. S13: enlarged XRD pattern ofFigure 2(b). Fig. S14: AFM images of the IE (a) and control(b) precursor films. Fig. S15: the PL spectra of the controlprecursor film collected from the perovskite side and theglass side. Fig. S16: the PL spectra of the annealed IE film col-lected from the perovskite side and the glass side. Fig. S17: theUV-vis spectral evolution of the control films during theannealing process. Fig. S18: XRD pattern evolution of thecontrol precursor films with different annealing durations.Fig. S19: Fourier transform infrared spectroscopy (FT-IR)

9Research

spectra of the IE precursor film after antisolvent dipping. Fig.S20: XRD patterns of the IE precursor films with differentannealing procedures. Fig. S21: ToF-SIMS of the annealedcontrol and IE samples. Table S1: summary on device perfor-mance parameters of FA0.95Cs0.05PbI3 (both the IE and con-trol types)-based perovskite solar cells from 32 devices.(Supplementary Materials)

References

[1] N.-G. Park, M. Grätzel, T. Miyasaka, K. Zhu, and K. Emery,“Towards stable and commercially available perovskite solarcells,” Nature Energy, vol. 1, no. 11, article 16152, 2016.

[2] J.-P. Correa-Baena, M. Saliba, T. Buonassisi et al., “Promisesand challenges of perovskite solar cells,” Science, vol. 358,no. 6364, pp. 739–744, 2017.

[3] A. K. Jena, A. Kulkarni, and T. Miyasaka, “Halide perovskitephotovoltaics: background, status, and future prospects,”Chemical Reviews, vol. 119, no. 5, pp. 3036–3103, 2019.

[4] A. Kojima, K. Teshima, Y. Shirai, and T. Miyasaka, “Organo-metal halide perovskites as visible-light sensitizers for photo-voltaic cells,” Journal of the American Chemical Society,vol. 131, no. 17, pp. 6050-6051, 2009.

[5] H.-S. Kim, C. R. Lee, J. H. Im et al., “Lead iodide perovskitesensitized all-solid-state submicron thin film mesoscopic solarcell with efficiency exceeding 9%,” Scientific Reports, vol. 2,no. 1, 2012.

[6] Q. Jiang, Y. Zhao, X. Zhang et al., “Surface passivation ofperovskite film for efficient solar cells,” Nature Photonics,vol. 13, no. 7, pp. 460–466, 2019.

[7] S. Pang, H. Hu, J. Zhang et al., “NH2CH═NH2PbI3: an alter-native organolead iodide perovskite sensitizer for mesoscopicsolar cells,” Chemistry of Materials, vol. 26, no. 3, pp. 1485–1491, 2014.

[8] N. J. Jeon, J. H. Noh, W. S. Yang et al., “Compositional engi-neering of perovskite materials for high-performance solarcells,” Nature, vol. 517, no. 7535, pp. 476–480, 2015.

[9] A. Amat, E. Mosconi, E. Ronca et al., “Cation-induced band-gap tuning in organohalide perovskites: interplay of spin–orbitcoupling and octahedra tilting,” Nano Letters, vol. 14, no. 6,pp. 3608–3616, 2014.

[10] C. Quarti, E. Mosconi, J. M. Ball et al., “Structural and opticalproperties of methylammonium lead iodide across the tetrag-onal to cubic phase transition: implications for perovskite solarcells,” Energy & Environmental Science, vol. 9, no. 1, pp. 155–163, 2016.

[11] E. J. Juarez-Perez, Z. Hawash, S. R. Raga, L. K. Ono, and Y. Qi,“Thermal degradation of CH3NH3PbI3 perovskite into NH3and CH3I gases observed by coupled thermogravimetry-massspectrometry analysis,” Energy & Environmental Science,vol. 9, no. 11, pp. 3406–3410, 2016.

[12] Z. Li, M. Yang, J. S. Park, S. H. Wei, J. J. Berry, and K. Zhu,“Stabilizing perovskite structures by tuning tolerance factor:formation of formamidinium and cesium lead iodide solid-state alloys,” Chemistry of Materials, vol. 28, no. 1, pp. 284–292, 2016.

[13] D. P. McMeekin, G. Sadoughi, W. Rehman et al., “A mixed-cation lead mixed-halide perovskite absorber for tandem solarcells,” Science, vol. 351, no. 6269, pp. 151–155, 2016.

[14] L. Xie, K. Lin, J. Lu et al., “Efficient and stable low-bandgapperovskite solar cells enabled by a CsPbBr3-cluster assisted

bottom-up crystallization approach,” Journal of the AmericanChemical Society, vol. 141, no. 51, pp. 20537–20546, 2019.

[15] M. Saliba, T. Matsui, J. Y. Seo et al., “Cesium-containing triplecation perovskite solar cells: improved stability, reproducibilityand high efficiency,” Energy & Environmental Science, vol. 9,no. 6, pp. 1989–1997, 2016.

[16] E. T. Hoke, D. J. Slotcavage, E. R. Dohner, A. R. Bowring, H. I.Karunadasa, and M. D. McGehee, “Reversible photo-inducedtrap formation in mixed-halide hybrid perovskites for photo-voltaics,” Chemical Science, vol. 6, no. 1, pp. 613–617, 2015.

[17] H. Min, M. Kim, S. U. Lee et al., “Efficient, stable solar cells byusing inherent bandgap of α-phase formamidinium leadiodide,” Science, vol. 366, no. 6466, pp. 749–753, 2019.

[18] S.-H. Turren-Cruz, A. Hagfeldt, andM. Saliba, “Methylammo-nium-free, high-performance and stable perovskite solar cellson a planar architecture,” Science, vol. 362, no. 6413,pp. 449–453, 2018.

[19] J.-W. Lee, H.-S. Kim, and N.-G. Park, “Lewis acid–base adductapproach for high efficiency perovskite solar cells,” Accounts ofChemical Research, vol. 49, no. 2, pp. 311–319, 2016.

[20] J.-W. Lee, Z. Dai, C. Lee et al., “Tuning molecular interactionsfor highly reproducible and efficient formamidinium perov-skite solar cells via adduct approach,” Journal of the AmericanChemical Society, vol. 140, no. 20, pp. 6317–6324, 2018.

[21] P. You, G. Li, G. Tang, J. Cao, and F. Yan, “Ultrafast laser-annealing of perovskite films for efficient perovskite solarcells,” Energy & Environmental Science, vol. 13, no. 4,pp. 1187–1196, 2020.

[22] J. Troughton, M. J. Carnie, M. L. Davies et al., “Photonic flash-annealing of lead halide perovskite solar cells in 1 ms,” Journalof Materials Chemistry A, vol. 4, no. 9, pp. 3471–3476, 2016.

[23] V. L. Pool, B. Dou, D. G. van Campen et al., “Thermal engi-neering of FAPbI3 perovskite material via radiative thermalannealing and in situ XRD,” Nature Communications, vol. 8,no. 1, article 14075, 2017.

[24] G. Li, T. Zhang, F. Xu, and Y. Zhao, “A facile deposition oflarge grain and phase pure α-FAPbI3 for perovskite solar cellsvia a flash crystallization,” Materials Today Energy, vol. 5,pp. 293–298, 2017.

[25] N. Rolston, R. Bennett-Kennett, L. T. Schelhas et al., “Com-ment on “Light-induced lattice expansion leads to high-efficiency perovskite solar cells”,” Science, vol. 368, no. 6488,article eaay8691, 2020.

[26] T. Liu, Y. Zong, Y. Zhou et al., “High-performanceformamidinium-based perovskite solar cells viamicrostructure-mediated δ-to-α-phase transformation,”Chemistry of Materials, vol. 29, no. 7, pp. 3246–3250, 2017.

[27] J.-W. Lee, D. H. Kim, H. S. Kim, S. W. Seo, S. M. Cho, andN. G. Park, “Formamidinium and cesium hybridization forphoto- and moisture-stable perovskite solar cell,” AdvancedEnergy Materials, vol. 5, no. 20, article 1501310, 2015.

[28] D. Yang, R. Yang, K. Wang et al., “High efficiency planar-typeperovskite solar cells with negligible hysteresis using EDTA-complexed SnO2,” Nature Communications, vol. 9, no. 1,p. 3239, 2018.

[29] D. H. Kim, J. Park, Z. Li et al., “300% enhancement of carriermobility in uniaxial-oriented perovskite films formed bytopotactic-oriented attachment,” Advanced Materials, vol. 29,no. 23, p. 1606831, 2017.

[30] F. Xie, C. C. Chen, Y. Wu et al., “Vertical recrystallization forhighly efficient and stable formamidinium-based inverted-

10 Research

structure perovskite solar cells,” Energy & Environmental Sci-ence, vol. 10, no. 9, pp. 1942–1949, 2017.

[31] G. Zheng, C. Zhu, J. Ma et al., “Manipulation of facet orienta-tion in hybrid perovskite polycrystalline films by cation cas-cade,” Nature Communications, vol. 9, no. 1, p. 2793, 2018.

[32] G. E. Eperon, S. D. Stranks, C. Menelaou, M. B. Johnston, L. M.Herz, and H. J. Snaith, “Formamidinium lead trihalide: abroadly tunable perovskite for efficient planar heterojunctionsolar cells,” Energy & Environmental Science, vol. 7, no. 3,pp. 982–988, 2014.

[33] H. Qifeng, S.-H. Bae, P. Sun et al., “Single crystal formamidi-nium lead iodide (FAPbI 3): insight into the structural, optical,and electrical properties,” Advanced Materials, vol. 28,pp. 2253–2258, 2016.

[34] L.-Q. Xie, L. Chen, Z. A. Nan et al., “Understanding the cubicphase stabilization and crystallization kinetics in mixed cationsand halides perovskite single crystals,” Journal of the AmericanChemical Society, vol. 139, no. 9, pp. 3320–3323, 2017.

[35] L. Chen, Y. Y. Tan, Z. X. Chen et al., “Toward long-term stabil-ity: single-crystal alloys of cesium-containing mixed cationand mixed halide perovskite,” Journal of the American Chem-ical Society, vol. 141, no. 4, pp. 1665–1671, 2019.

[36] Y. Lin, Y. Fang, J. Zhao et al., “Unveiling the operation mech-anism of layered perovskite solar cells,” Nature Communica-tions, vol. 10, no. 1, p. 1008, 2019.

[37] C. Yi, J. Luo, S. Meloni et al., “Entropic stabilization of mixedA-cation ABX3metal halide perovskites for high performanceperovskite solar cells,” Energy & Environmental Science,vol. 9, no. 2, pp. 656–662, 2016.

[38] M. Kim, G. H. Kim, T. K. Lee et al., “Methylammonium chlo-ride induces intermediate phase stabilization for efficientperovskite solar cells,” Joule, vol. 3, no. 9, pp. 2179–2192, 2019.

[39] N. J. Jeon, J. H. Noh, Y. C. Kim, W. S. Yang, S. Ryu, and S. I.Seok, “Solvent engineering for high-performance inorganic-organic hybrid perovskite solar cells,” Nature Materials,vol. 13, no. 9, pp. 897–903, 2014.

[40] S. Shao, J. Liu, G. Portale et al., “Highly reproducible Sn-basedhybrid perovskite solar cells with 9% efficiency,” AdvancedEnergy Materials, vol. 8, no. 4, article 1702019, 2018.

[41] L. Ma, D. Guo, M. Li et al., “Temperature-dependent thermaldecomposition pathway of organic–inorganic halide perov-skite materials,” Chemistry of Materials, vol. 31, no. 20,pp. 8515–8522, 2019.

[42] C. Li, Y. Zhou, L. Wang et al., “Methylammonium-mediatedevolution of mixed-organic-cation perovskite thin films: adynamic composition-tuning process,” Angewandte Chemie,International Edition, vol. 56, no. 26, pp. 7674–7678, 2017.

[43] J. Burschka, N. Pellet, S. J. Moon et al., “Sequential depositionas a route to high-performance perovskite-sensitized solarcells,” Nature, vol. 499, no. 7458, pp. 316–319, 2013.

[44] Y.Wu, A. Islam, X. Yang et al., “Retarding the crystallization ofPbI2for highly reproducible planar-structured perovskite solarcells via sequential deposition,” Energy & Environmental Sci-ence, vol. 7, no. 9, pp. 2934–2938, 2014.

[45] W. S. Yang, J. H. Noh, N. J. Jeon et al., “High-performancephotovoltaic perovskite layers fabricated through intramolecu-lar exchange,” Science, vol. 348, no. 6240, pp. 1234–1237, 2015.

[46] T. Zhang, M. Yang, Y. Zhao, and K. Zhu, “Controllablesequential deposition of planar CH3NH3PbI3 perovskite filmsvia adjustable volume expansion,” Nano Letters, vol. 15, no. 6,pp. 3959–3963, 2015.

[47] Y. Zhao, H. Tan, H. Yuan et al., “Perovskite seeding growth offormamidinium-lead-iodide-based perovskites for efficientand stable solar cells,” Nature Communications, vol. 9, no. 1,article 1607, 2018.

[48] B. Chen, Z. Yu, K. Liu et al., “Grain engineering for perovski-te/silicon monolithic tandem solar cells with efficiency of25.4%,” Joule, vol. 3, no. 1, pp. 177–190, 2019.

[49] C. Zhu, X. Niu, Y. Fu et al., “Strain engineering in perovskitesolar cells and its impacts on carrier dynamics,” Nature Com-munications, vol. 10, no. 1, p. 815, 2019.

[50] J. Zhao, Y. Deng, H. Wei et al., “Strained hybrid perovskitethin films and their impact on the intrinsic stability of perov-skite solar cells,” Science Advances, vol. 3, no. 11, 2017.

[51] M. Wei, K. Xiao, G. Walters et al., “Combining efficiency andstability in mixed tin–lead perovskite solar cells by cappinggrains with an ultrathin 2D layer,” Advanced Materials,vol. 32, no. 12, article 1907058, 2020.

[52] L. Meng, C. Sun, R. Wang et al., “Tailored phase conversionunder conjugated polymer enables thermally stable perovskitesolar cells with efficiency exceeding 21%,” Journal of the Amer-ican Chemical Society, vol. 140, no. 49, pp. 17255–17262, 2018.

11Research